BeforeClusterUpgrade is not called with cp unavailable

### What steps did you take and what happened?

Using `ClusterClass` we found an unexpected behavior when we try to upgrade the cluster and intercept the `BeforeClusterUpgrade` hook using RuntimeSDK.
If a ControlPlane isn't in an healthy state the hook doesn't get propagated to the runtime extension, I think this is logically wrong and in contrast with the [documentation](https://cluster-api.sigs.k8s.io/tasks/experimental-features/runtime-sdk/implement-lifecycle-hooks.html?highlight=beforeclusterup#beforeclusterupgrade).

### What did you expect to happen?

To receive the `BeforeClusterUpgrade` request even if the ControlPlanes are unhealthy and manage the problem in the phase of upgrading.

### Cluster API version

1.4.2+

### Kubernetes version

1.27.2

### Anything else you would like to add?

_No response_

### Label(s) to be applied

/kind bug
One or more /area label. See https://github.com/kubernetes-sigs/cluster-api/labels?q=area for the list of labels.

Start by reading the RuntimeSDK documentation for the BeforeClusterUpgrade hook and trace the ClusterClass upgrade path where ControlPlane health is checked. Reproduce the unhealthy-ControlPlane case, then verify that the hook request reaches the runtime extension and that the existing upgrade behavior remains intact.

Written by the indexing model from the issue text.